Leveraging machine learning techniques and SEVIRI data to detect volcanic clouds composed of ash, ice, and SO ₂

Abstract. Volcanic clouds can influence the climate and pose a serious threat to air transportation. Detecting and distinguishing them from meteorological clouds is particularly challenging because they often are composed of water vapor and ice particles, along with ash and gases. This study presents a neural network (NN) model for the detection of volcanic clouds composed of ash, ice, and SO2, applied to data acquired by the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) satellite instrument. A dataset of 1259 SEVIRI images related to Mount Etna volcano (Italy) eruptions spanning from 2020 to 2022, as well as 2024, was considered. The NN model, based on a multi-layer perceptron (MLP), was developed using 13 features, including thermal infrared channels and brightness temperature differences (BTDs). A post-processing step based on a plume-tracking algorithm and a Non-Local means filter was implemented to improve the performance of the NN model. The model was validated using three eruptive events that were not included in the training phase, achieving an overall balanced accuracy of up to 92.0 %. The validation results also showed that the model successfully detected 66.0 %, 48.5 %, and 84.1 % of the observed volcanic cloud (VC) pixels in the three analysed validation events, respectively. In addition, only 7.7 %, 4.0 %, and 21.9 % of the detected VC pixels corresponded to false alarms for the respective events. Thus, the model demonstrates the capability to detect volcanic clouds even under complex conditions of high meteorological cloud cover. The results are promising for the automatic detection of volcanic clouds, including those containing ice and SO2, as well as for improving volcanic cloud retrieval processes.